December, 6th-7th 2016

Prologue

  • doing a lot of exploration in RNA-seq data
  • lack of something to do this interactively
  • even better, fully powered by the Bioconductor ecosystem

pcaExplorer came to life (Bioc 3.3)

  • empower the domain expert (extensively used by our coop partners)
  • transparency, internal & independent validation, help for future self was also needed!

The next challenge

Interactive and reproducible research \(\rightarrow\) analyze, visualize, integrate

  • Scripts are great but hard
  • GUIs are quite fun but not so flexible or transparent
  • No biologist will escape computational analysis - pick your poison!

Extending the original recipe

  • RNA-seq data + R & Shiny + PCA

  • little great help from literate programming (Rmarkdown)
  • SPSS envy: point & click but also logging

Version 2.0 includes the newest developments

Ready for a quick tour?

Interactivity: data overview

Interactivity: exploration of PCs (samples)

Interactivity: exploration of PCs (genes)

Interactivity: counts table exporting

Interactivity: identification of outlier samples

Interactivity: quick peek on shortlisted genes

Interactivity: functional annotation of PCs

Reproducibility: editing & generation of report

Reproducibility: editing & generation of report

Reproducibility: state saving & store workspace

  • fully integrated in Bioconductor –> –>

The design choices in detail

  • embedded text editor: exploiting shinyAce
  • no white sheet syndrome: provide template report, editable if needed - plus, include a preview in the app
  • reporting: Rmarkdown + reactiveValues accessed when knitting and rendering
  • saving & sharing: binary objects to workspace/global environment (wait for Shiny support for bookmarking?)

Very happy to include your suggestions!

All in all…

Interactivity and reproducibility can be efficiently combined together!



  • make data exploration accessible and reproducible + extendable approach
  • quick & comprehensive report generation, plus state saving to share workspace
  • real practical companion for any RNA-seq dataset, fully integrated in Bioconductor (http://bioconductor.org/packages/pcaExplorer/)

Coming soon: ideal

Coming soon: ideal

Coming soon: ideal

Outlook

  • pcaExplorer to include more methods for dimension reduction
  • the ideal way of doing DE analysis is about to come
  • new features to be implemented: refinement in the downstream analysis, support for quasi-mapping approaches (a la tximport)
  • mind the gap - and bridge it: deployment of the apps also in standalone Shiny server


Development of Applications for Interactive and Reproducible Research: a Case Study - Federico Marini and Harald Binder (latest issue of Genomics Computational Biology)

pcaExplorer: an R/Bioconductor package for interacting with RNA-seq principal components - Federico Marini and Harald Binder (submitted)

Acknowledgements

  • IMBEI - Bioinformatics division
    • Harald Binder and all the colleagues for inputs
  • Center for Thrombosis and Hemostasis (CTH), Mainz
    • Wolfram Ruf
    • Sebastian Schubert
    • … and many other precious data generators